Traditional indoor pedestrian positioning technology based on magnetic field information determines the user’s location by matching the magnetic field fingerprint database collected during the offline phase with the magnetic field intensity measurements obtained during the online phase. However, in practical application environments, influenced by complex factors, such as magnetometer heterogeneity and temporal variations, the magnetic field information collected in these two phases often leads to significant discrepancies. The greater the difference between the magnetic field intensity stored in the fingerprint database and that collected by the user, the lower the positioning accuracy. This study proposes a cascaded filtering algorithm to address the precondition constraints of an inertial navigation system. In the lower filter, zero-velocity correction and Attitude Extended Complementary Filtering (ECF) are utilized to initially solve the pedestrian trajectory. In the upper filter, both the magnetic field intensity and magnetic induction ratio are used for the dual calibration of the accumulated errors during the positioning process. By performing an inverse calculation based on the magnetic matching results, the estimated initial position and initial heading angle are derived, thereby overcoming the limitations of inertial navigation in indoor positioning. To evaluate the performance of the proposed technique, experiments were conducted along a 98-meter test path. The results show that the proposed algorithm achieves a root mean square error (RMSE) of only 0.472 m, demonstrating superior positioning accuracy compared with the current state-of-the-art in magnetic field-assisted positioning research.

Open-set test-time adaptation (OSTTA) addresses the challenge of adapting models to new environments where out-of-distribution (OOD) samples coexist with in-distribution (ID) samples affected by distribution shifts. In such settings, covariate shift-for example, changes in weather conditions such as snow-can alter ID samples, reducing model reliability. Consequently, models must not only correctly classify covariate-shifted ID (csID) samples but also effectively reject covariate-shifted OOD (csOOD) samples. Entropy minimization is a common strategy in test-time adaptation to maintain ID performance under distribution shifts, while entropy maximization is widely applied to enhance OOD detection. Several studies have sought to combine these objectives to tackle the challenges of OSTTA. However, the intrinsic conflict between entropy minimization and maximization inevitably leads to a trade-off between csID classification and csOOD detection. In this paper, we first analyze the limitations of entropy maximization in OSTTA and then introduce an angular loss to regulate feature norm magnitudes, along with a feature-norm loss to suppress csOOD logits, thereby improving OOD detection. These objectives form ROSETTA, a robust open-set test-time adaptation. Our method achieves strong OOD detection while maintaining high ID classification performance on CIFAR-10-C, CIFAR-100-C, Tiny-ImageNet-C and ImageNet-C. Furthermore, experiments on the Cityscapes validate the method's effectiveness in real-world semantic segmentation, and results on the HAC dataset demonstrate its applicability across different open-set TTA setups.

Accurate and continuous recognition of individual animal behaviors is fundamental to intelligent livestock farming and precision management. Traditional manual inspection lacks scalability and real-time capability, while vision-based approaches are often constrained by occlusion, illumination variation, and complex barn environments. Wearable inertial sensors provide a robust alternative for individual-level behavior perception with weak dependence on environmental conditions. However, pig behaviors exhibit pronounced heterogeneity in temporal scales and motion patterns, making accurate temporal modeling under limited computational resources particularly challenging for edge deployment. To address these challenges, this paper proposes a Multi-scale Semantic-aware Temporal Network (MSATNet) for sow behavior recognition using wearable six-axis inertial sensors. MSATNet combines a Large-scale Temporal Modeling (LTM) module and a Small-scale Temporal Refinement (STR) module to jointly model long-duration stable behaviors and short-duration transient motions within fixed time windows. In addition, a Semantic-aware Supervision (SAS) mechanism is introduced during training to incorporate hierarchical behavioral semantics and reduce confusion among semantically similar behaviors without increasing inference cost. The proposed method was evaluated in a real-world large-scale pig farm. MSATNet achieves accuracies of 93.69% on the validation set and 92.19% on an independent test set across seven daily sow behaviors, outperforming Random Forest (88.59%), MobileNet V2 (89.59%) and ViT (91.99%) under identical experimental settings. Furthermore, it achieves a single-sample inference latency of 1.34 ms on the Raspberry Pi 5 edge device, demonstrating its suitability for real-time deployment. These results indicate that MSATNet provides an accurate, lightweight, and scalable solution for continuous pig behavior monitoring in edge-intelligent livestock farming systems.

Electron tomography (ET) is crucial for determining the three-dimensional (3D) structure of materials in real space but challenging due to the inherent missing wedge, high dose, and limited depth-of-field. Although deep learning can address these challenges in principle, the scarcity of ET data severely limits its application. In this study, we propose a general data-driven ET reconstruction framework that uses extensive and readily available random high-entropy projections to construct large-scale datasets. By integrating both real structural priors and depth-dependent imaging physics, the framework enables high-quality 3D reconstruction independent of specific materials or resolutions. Using this strategy, we successfully determine the 3D atomic structure of a 13-nm Pt nanoparticle containing 52 138 atoms, achieving a root-mean-square displacement of 22.6 pm; the projection consistency error is significantly reduced, effectively expanding the depth-of-field limit of atomic-scale ET.

In low-frequency vibrating MEMS resonators, device structures typically vibrate in flexural mode and torsional mode. The mechanical stiffness in the motion direction of the bending mode is often strongly correlated with the device’s equivalent stiffness, while torsional mode generally use piezoelectric actuation and usually have a low Q factor. These factors limit the application of low-frequency resonators in some high-reliability scenarios. This paper proposes a low-frequency wheel-shaped resonator vibrating in the in-plane torsional mode. By using the torsional vibration of the wheel-shaped mass instead of translational vibration, and adopting an in-plane capacitive structure for driving and sensing, the device maintains high in-plane stiffness while vibrating at low frequencies. At the same frequency, its in-plane stiffness is 10² orders of magnitude higher than that of the flexural mode. Simulations on the anti-overload performance of the device structure show that under a 100g static load, the frequency drifts in the X-axis and Z-axis directions are 21.67 ppb and 60.39 ppb, respectively. Under 10000g impact loads in the X-axis and Z-axis directions, the maximum in-plane stresses are 24.7 MPa and 45.7 MPa, with maximum displacements of 0.111 µm and 0.137 µm, respectively—values far lower than the maximum allowable stress of silicon material and the minimum line width of the device. This demonstrates that the structure has high frequency stability and anti-overload performance.

Low-cost, high-precision indoor positioning systems are becoming crucial in some applications such as smart manufacturing, logistics and warehousing, pedestrian tracking and embodied artificial intelligence. Ultrawideband (UWB)-based ranging techniques offer advantages like short pulse intervals and high temporal resolution, but non-line-of-sight (NLOS) occlusion can limit their performance and application scope. Low-cost inertial measurement units (IMUs) provide accurate navigation information over short periods but suffer from error accumulation. To address these issues, this paper propose a IMU/UWB tightly coupled navigation algorithm combining carrier motion characteristics. In line-of-sight (LOS) environments, the precise positioning information from UWB assists in correcting the accumulated error of IMU. In NLOS environments, the corrected inertial navigation system (INS) information improves the system's robustness and accuracy. Additionally, we incorporate the carrier motion information into the INS using the extended Kalman filter algorithm and use the results for NLOS determination. The experimental results show that the tight coupling algorithm proposed in this paper can effectively eliminate the accumulated error of IMU. Under LOS conditions, the root mean square error is reduced by 29.9% compared with single UWB positioning. At the same time, the large-scale NLOS problem is solved while ensuring positioning accuracy. This method can reduce the deployment density of base stations in practical applications, expand the effective positioning range of the system, and make positioning in some NLOS areas possible.

This work aims to introduce for the first time the concept of "in situ release" for MEMS sensors and validate the feasibility of this technology using an accelerometer structure with a modified design. MEMS sensors enhanced with in situ release technology will be able to fix their movable structures in place using thermally decomposable materials during the fabrication process and can be released as needed during their operational phase on boards thereby enhancing the ability of the fragile MEMS structures to withstand acceleration overload events. This research covers the entire process of the in situ release technology including the selection and formulation of thermally decomposable materials, the development of a precise droplet addition system for the materials, the design and fabrication of a verification structure with microheaters, and the final testing phase. Driven by voltage signals, the microheaters functioned effectively, enabling the efficient decomposition of the thermally decomposable material and the restoration of the structure's mobility. The decomposition process was documented and analyzed, thereby validating the feasibility of the in situ release technology.

Human activity recognition (HAR) is an application of great importance in the Internet of Things (IoT). Inertial measurement units (IMU) on wearable devices provide the primary source of time-series data for HAR. This paper focuses on addressing the energy consumption and performance issues in real-world scenarios for continuous HAR using time-series signals. We present a continuous adaptive spiking neural network (CASNN) suitable for low-power wearable devices. CASNN is implemented by introducing early exit into SNN, and the early exit branches do not require additional classifiers. The results show that CASNN can reduce over 56% FLOPs and improve accuracy by 4% on average across two datasets, through additional cross-person generalization capabilities on the continuous HAR dataset.

With the continuous development of the IoT, compact wireless communication modules have become indispensable components, and their antennas are gradually being developed from external devices into onboard integrated devices. The serpentine antenna, a variant of the monopole antenna known for its small size and easy integration, is often applied to engineering practices. However, its performance has always been closely affected by the size of the surrounding grounding plane. By conducting a characteristic mode analysis (CMA), this study explored the variation patterns in the ground plane size and the resonant frequency. Based on the simulation results, it was clear that when the ground plane size is less than a quarter of the working wavelength, the ground plane will have a significant effect on the antenna's resonant frequency. Thus, this study further analyzed a serpentine antenna with a grounding branch, and through analysis of the basic law of the influence of grounding structure on the antenna's performance, we found that by adjusting the branch length, the matching performance of the antenna can be effectively improved. Furthermore, by changing the size of the ground plate, the antenna's resonant frequency can be adjusted. Such a conclusion will hopefully provide a reference for future designs of integrated antennas in engineering applications.

Robust and accurate attitude and heading estimation using Micro-Electromechanical System (MEMS) Inertial Measurement Units (IMU) is the most crucial technique that determines the accuracy of various downstream applications, especially pedestrian dead reckoning (PDR), human motion tracking, and Micro Aerial Vehicles (MAVs). However, the accuracy of the Attitude and Heading Reference System (AHRS) is often compromised by the noisy nature of low-cost MEMS-IMUs, dynamic motion-induced large external acceleration, and ubiquitous magnetic disturbance. To address these challenges, we propose a novel data-driven IMU calibration model that employs Temporal Convolutional Networks (TCNs) to model random errors and disturbance terms, providing denoised sensor data. For sensor fusion, we use an open-loop and decoupled version of the Extended Complementary Filter (ECF) to provide accurate and robust attitude estimation. Our proposed method is systematically evaluated using three public datasets, TUM VI, EuRoC MAV, and OxIOD, with different IMU devices, hardware platforms, motion modes, and environmental conditions; and it outperforms the advanced baseline data-driven methods and complementary filter on two metrics, namely absolute attitude error and absolute yaw error, by more than 23.4% and 23.9%. The generalization experiment results demonstrate the robustness of our model on different devices and using patterns.

The main temperature compensation method for MEMS piezoresistive pressure sensors is software compensation, which processes the sensor data using various algorithms to improve the output accuracy. However, there are few algorithms designed for sensors with specific ranges, most of which ignore the operating characteristics of the sensors themselves. In this paper, we propose three temperature compensation methods based on swarm optimization algorithms fused with machine learning for three different ranges of sensors and explore the partitioning ratio of the calibration dataset on Sensor A. The results show that different algorithms are suitable for pressure sensors of different ranges. An optimal compensation effect was achieved on Sensor A when the splitting ratio was 33.3%, where the zero-drift coefficient was 2.88 × 10−7/°C and the sensitivity temperature coefficient was 4.52 × 10−6/°C. The algorithms were compared with other algorithms in the literature to verify their superiority. The optimal segmentation ratio obtained from the experimental investigation is consistent with the sensor operating temperature interval and exhibits a strong innovation.

High Energy Electron Radiography (HEER) has been proposed as a new material diagnostic technology in recent years. The main features of this technology are the strong penetrating power, high space-time resolution, and large area density diagnostic range. Therefore, it is considered as one of the effective diagnostic methods in the field of high energy density material diagnosis. For further research of HEER, the High Energy Electron Radiography Platform in Lanzhou (HERPL) has designed and built as a dedicated experimental platform of HEER, which is mainly composed of a 50MeV electron linear accelerator based on a thermionic cathode RF gun, and a set of quadrupole magnet image systems. In this paper, the HEER experiment has completed and the spatial resolution of 1μm, the density resolution of 1% was obtained which reached the world record of HEER.

MEMS device degradation due to aging and other factors is becoming a major concern because it will cause parametric deviations and catastrophic failures in the mechanical and structural subsystems. However, MEMS testing in general which needs specific sophisticated testing equipment is complicated and time-consuming. To solve these problems, this paper specifically introduces a built-in self-test method which based on the periodic observation of the temperature-dependent output signal. A packaging scheme is designed and the test circuit is built to conduct test experiments on MEMS pressure sensors with different ranges and materials. The experimental results show that this method can effectively test the performance and will not affect the continued normal operation of the sensor. Furthermore, some compensation is made to correct the output to greatly improve the accuracy and reliability. Low cost, ease of implementation, and possibility to monitor in real time are the main advantages.

Piezoresistive pressure sensors have been widely used in the industry and many other fields. However, conventional pressure sensor inspection is performed with the device off, which causes several inconveniences. In this article, an electrothermal actuator is designed and fabricated on the basis of bimetallic alloy material to realise the self-test of the piezoresistive silicon pressure sensor, and a kind of pressure sensor package structure with self-test capability is developed using the electrothermal actuator. Experimental results show that the structure can realise the dual excitation of the pressure and temperature of the diffused silicon pressure sensor core and enable it to return to the normal working state in less than 3.3 s. The self-test module can be driven at a low voltage of 0.8 V at less than 618 mW and can reflect the performance of the pressure sensor quickly and accurately. The structure is suitable for most differential pressure, absolute pressure sensor chips.

Indoor pedestrian positioning has been widely used in many scenarios, such as fire rescue and indoor path planning. Compared with other technologies, inertial measurement unit (IMU)-based indoor positioning requires no additional equipment and has a lower cost. However, IMU-based indoor positioning has the problem of error accumulation, resulting in inaccurate positioning. Therefore, this paper proposes a cascade filtering algorithm to correct the accumulated error using only a small amount of map information. In the lower filter, the zero-velocity correction and the attitude-extended complementary filtering (ECF) algorithm are utilized to initially solve the pedestrian’s trajectory. In the upper filter, a particle filter (PF) combined with the map information is adopted to correct the accumulated error of the heading and stride length. In the 2D positioning process, the root mean square error (RMSE) of the proposed algorithm is only 1.35 m. In the altitude correction, this paper proposes a method of clustering floor discrimination to deal with the instability of the barometer resulting from an uneven pressure and temperature. In the final 3D positioning experiment, with a total length of 536.5 m and including the process of going up and down the stairs, the end-point error is only 2.45 m by the proposed algorithm.

Robust and accurate human stride length estimation (SLE) using smartphone integrated inertial measurement units (IMU) is essential in pedestrian dead reckoning (PDR) and mobile health applications. However, the change of smartphone carrying mode (i.e. sensor location) in daily usage often leads to significant estimation errors. To address this problem, we propose a novel SLE framework called Mode-Independent Neural Network (MINN) using multi-source unsupervised domain adaptation (UDA) methods. First, we present a hierarchical neural network to extract spatio-temporal features based on multi-level ResNet and GRU. Then, we use adversarial training and a subclass classifier to build a UDA network that can extract mode-invariant features shared by the data from different modes. Finally, we integrate these architectures into an end-to-end learning framework. Through a systematic evaluation under the leave-one-out setting on two public SLE datasets, the MINN outperforms the state-of-the-art algorithms by achieving stride length error rates of 2.5% and 5.1% in supervised settings. We also evaluated the mode-independent adaptability of this model by performing single and multiple UDA tasks. The results demonstrate that the proposed MINN significantly improves the generalization of SLE model under new subjects or modes.

With the rapid development of Internet of Things (IoT) technologies, traditional disease diagnoses carried out in medical institutions can now be performed remotely at home or even ambient environments, yielding the concept of the Internet of Health Things (IoHT). Among the diverse IoHT applications, inertial measurement unit (IMU)-based systems play a significant role in the detection of diseases in many fields, such as neurological, musculoskeletal, and mental. However, traditional numerical interpretation methods have proven to be challenging to provide satisfying detection accuracies owing to the low quality of raw data, especially under strong electromagnetic interference (EMI). To address this issue, in recent years, machine learning (ML)-based techniques have been proposed to smartly map IMU-captured data on disease detection and progress. After a decade of development, the combination of IMUs and ML algorithms for assistive disease diagnosis has become a hot topic, with an increasing number of studies reported yearly. A systematic search was conducted in four databases covering the aforementioned topic for articles published in the past six years. Eighty-one articles were included and discussed concerning two aspects: different ML techniques and application scenarios. This review yielded the conclusion that, with the help of ML technology, IMUs can serve as a crucial element in disease diagnosis, severity assessment, characteristic estimation, and monitoring during the rehabilitation process. Furthermore, it summarizes the state-of-the-art, analyzes challenges, and provides foreseeable future trends for developing IMU-ML systems for IoHT.

A CW mode RF modulated grid-controlled thermionic electron gun was proposed by Institute of Modern Physics (IMP), Chinese Academy of Sciences (CAS) for some high average current electron accelerators requirements. The RF modulated grid-controlled thermionic electron gun was selected for these purposes due to its simplicity and cost savings. The experimental proof test of this type electron gun was conducted. The RF power supply at 107.5 MHz for the grid modulation can be adjusted from 10 W to 70 W. The RF power is coupled into the gun of grid-cathode through a RF&DC modulator. The electromagnetic field and RF simulation of the modulator is presented here. The gun structure and the beam dynamics design are also shown in the paper. The cathode assembly and the electron guns are tested on a 10 kV test bench for beam characterization. The CW mode 107.5 MHz electron beam obtained from the proof test, which is important for future high average current electron injectors development.

为实现微机电系统(micro-electro-mechanical system,MEMS)加速度计在应用过程中的实时补偿校准,以保证应用需求的高精度输出,本研究在建立测量值与真实值之间的加速度自校准模型的基础上,对加速度计在任意位置下的多组静态观测数据样本进行筛选,结合LM(levenberg-marquardt)算法和最小二乘法模型参数,优化了LM算法中过度依赖初值的问题.对于任意位置下的加速度计静态输出数据,滤波后筛选出可用于最小二乘法的姿态数据,用来修正部分或者全部第k次迭代模型参数,作为第k+1次迭代的初值;其他姿态数据用于LM算法训练拟合第k+1次迭代模型参数,实现加速度计应用过程中的闭环、实时校准.以智能鞋垫应用为例,本研究对比了传统十二面体法、椭球法、单纯LM算法和LM&最小二乘法自校准法对加速度计的校准结果.结果表明,在智能鞋垫的长期使用中,本研究提出的LM&最小二乘法自校准法消除了由于LM初值设定引起的模型参数解算不精准的情况,并可实现实时采集、实时解算、实时校准的目标,能够达到与传统标定方法相同量级的姿态精度.

As a new scheme, High Energy Electron Radiography (HEER) was considered as an effective diagnostic tool in the mesoscale sciences due to its high spatial temporal resolution and large area density diagnostic range. Some improvements were implemented to achieve high spatial resolution using a 50 MeV electron beam in Lanzhou, China. These included improving the uniformity of the beam transverse distribution, optimizing the energy spread and increasing the magnet lens magnification. Then the HEER image was collected by a CCD camera with image processing program and in-situ spatial resolution optimization adjustment program. To the best of our knowledge, the achieved spatial resolution of 0.8 mu m constitutes a world record. The details of this experiment are described in this publication.